Opened 10 years ago
Closed 10 years ago
#745 closed defect (fixed)
tao_test failing with quadruple precision
| Reported by: | Juergen Reuter | Owned by: | Juergen Reuter |
|---|---|---|---|
| Priority: | P0 | Milestone: | v2.2.8 |
| Component: | vamp | Version: | 2.2.7 |
| Severity: | normal | Keywords: | |
| Cc: |
Description
both for gfortran and ifort
Change History (5)
comment:1 Changed 10 years ago by
| Owner: | changed from ohl to Juergen Reuter |
|---|---|
| Priority: | P3 → P0 |
| Status: | new → assigned |
comment:2 Changed 10 years ago by
| Resolution: | → fixed |
|---|---|
| Status: | assigned → closed |
comment:4 Changed 10 years ago by
| Resolution: | fixed |
|---|---|
| Status: | closed → reopened |
comment:5 Changed 10 years ago by
| Resolution: | → fixed |
|---|---|
| Status: | reopened → closed |
Supplying the extra digits suggests an incorrect precision, since the RNG uses only 52bits internally. Fortunately, it suffices to specify an expected value the correct kind tao_r64 instead of default.
Fixed in r7267.
Note: See
TracTickets for help on using
tickets.
As it looks to me the failure of tao_test which was only in the test of the 52bit random numbers was only due to the fact that the reference value for the random number for Knuth's test, x(1009*2009+1) = x(2027083) had only been calculated up to 20 significant digits. That was ok for double and extended precision, but not for true quadruple precision which apparently needs 36 significant values. JRR inserted the more precise value and changed the Fortran output format in the test. This still has to be approved by TO. Closing for now. In case, TO (or anyone else) disagrees, please reopen.